Convex method for selection of fixed effects in high-dimensional linear mixed models

TL;DR

This paper introduces two convex methods for variable selection in high-dimensional linear mixed models, enabling the analysis of complex genetic and other data with many variables, outperforming existing non-convex approaches.

Contribution

The paper presents novel convex algorithms for variable selection in high-dimensional linear mixed models, expanding the capacity to handle more variables than previous non-convex methods.

Findings

Convex methods outperform non-convex methods in variable selection.

The methods effectively handle high-dimensional genetic data.

A proposed approach extends to a wider class of linear mixed models.

Abstract

Analysis of high-dimensional data is currently a popular field of research, thanks to many applications e.g. in genetics (DNA data in genomewide association studies), spectrometry or web analysis. At the same time, the type of problems that tend to arise in genetics can often be modelled using linear mixed models in conjunction with high-dimensional data because linear mixed models allow us to specify the covariance structure of the models. This enables us to capture relationships in data such as the population structure, family relatedness, etc. In this paper we introduce two new convex methods for variable selection in high-dimensional linear mixed models which, thanks to convexity, can handle many more variables than existing non-convex methods. Both methods are compared with existing methods and in the end we suggest an approach for a wider class of linear mixed models.